[mb-devel] Fwd: Acoustic fingerprinting

[Ben]

Well this of course is amazingly cool, if the guy thinks he can  
develop something that will cheaply scale to a billion songs.  
Especially if it sounds like it won't be patent-encumbered. Call me  
skeptical, but I'll believe it when I see it.

That said, I had put together 12GHz of CPUs to analyze fdmf, and I  
can keep them around to analyze any new algorithms that might pop up.  
Just send me the code.

On Oct 13, 2005, at 1:43 AM, Robert Kaye wrote:

> Here is a very interesting email I received from the former founder  
> of Tuneprint. Anyone who is interested in hacking on a TRM  
> replacement: YOU NEED TO READ THIS!
>
> Begin forwarded message:

[ SNIP ]

>> If you were interested in having me put some fingerprinting stuff
>> together, what I'd want from you and the MusicBrainz community is
>> testing and implementation support. That would mean taking some C  
>> code
>> from me that takes PCM and spits out a string and putting an mp3
>> decoder, network code, &c around it, and installing and maintaining a
>> server, which would also involve writing some scripts or Python  
>> code to
>> keep the database well-fed and complacent. It would also mean putting
>> together a group of volunteers with large collections willing to
>> download builds and run the tagger on their collection every few  
>> days,
>> and a few especially motivated volunteers willing to put together  
>> some
>> test scripts and tell me how well we're doing on accuracy. In other
>> words, all the things your organization already does very well :)